1.0.9

Changes from v1.0

• Updated manual
• Switch -Y introduced: use synonyms from a .csv file.
• Switch -U introduced: customizes the output, selecting different columns to be displayed.
• Multiple file input after switch --

1.0

Changes from v0.9.9

• Updated manual
• Switch -J introduced: It adds a column in the output with the confidence for superiority to the next player in the list.
• Format fixed for -j switch and elostat output.
• Relative priors between engines will not purge engines with perfect scores (or zero score)
• Code cleanup
• Increase precision of the function get_pWDL (which gets probabilities): Solved some pending issues.
• Introduce asserts to catch potential future precision issues
• When adjusting white advantage and draw rates, slowly increment deltas if the optimum is not at the center. Speeds up convergence of pathological cases.
• Extra warning for ill connected databases
• White advantage and draw rate simulated are based on the initial result, not from the previous simulation.
• In simulations, white advantage and draw rate are recalculated, so now errors for them are displayed
• Convergence speed increased by ~40% (varies according to input files).
• Two bugs fixed that delayed convergence in some pathological input data.
• White advantage and draw rate calculation re-designed and optimized to be more accurate and faster.
• Simulations checked for proper connectivity. Skipped if they are problematic.
• Switch -M introduced, which forces a calculation by Maximum likelihood criteria even if no loose anchors or priors are used. Slower than traditional, but serves as a comparison.
• Fix a convergence bug when a player with all wins is present
• Modify output, include decimal adjustments in -j output
• Groups connected output (-g) just a message when there is only one group
• Tweak parameters of convergence

Ordo goes parallel

• Introduction of switch -n [value], which sets number of cores used for parallel calculation during simulation

Ordo v0.9.9

Changes

• Number of players and games limited by the memory of the system
• Automatically detect isolated groups (as in the previously implemented -g switch). Issues a warning when problems are found.
• Engines are sorted alphabetically in the output of switch -g
• Convergence is forced to be reach a higher precision
• Internal sorting optimized
• Clean up of the code and minor bugs fixed. Extensive re-factoring 46 files changed, 5057 insertions(+), 3009 deletions(-)
• Manual updated

Switches introduced

• -t threshold of a minimum number of games played for a participant
• -G forces Ordo NOT to check for isolated groups in the database (it avoid warnings)